Federated learning method improves eye disease detection despite noisy labels
OCT-FedSIR: Toward Trustworthy Federated Ophthalmic Learning under Annotation Noise
Machine LearningArtificial IntelligenceDistributed, Parallel, and Cluster Computing
Summary
When many clinics work together to train eye disease detection models without sharing patient data, some may have errors in their labels, which makes training harder. The authors created OCT-FedSIR, a system that can spot and fix these labeling errors even when different clinics have different kinds of patients. This approach improves accuracy and keeps useful data rather than throwing it away. It was tested on multiple eye imaging datasets, showing better performance than other methods under various challenging conditions.
What this means in practice
- •For medical imaging teams: Improve federated training of eye disease detection models by identifying and correcting client-specific labeling errors without losing valuable data.
- •For federated learning engineers: Enhance robustness of federated classification systems against heterogeneous annotation noise by integrating spectral-based reliability checks and relabeling.
Authors
Sina Gholami, Abdulmoneam Ali, Tania Haghighi, Rashadul H. Badhon, Behafarin Emam, Sally S. Y. Ong, Atalie C. Thompson, Theodore Leng, Ahmed Arafa, Jennifer I. Lim, Minhaj Nur Alam
Abstract
Federated learning enables collaborative model development without centralizing patient data, but annotation reliability at participating institutions cannot always be assumed. In ophthalmic imaging, differences in disease prevalence and class composition can resemble changes caused by corrupted supervision. We introduce OCT-FedSIR, a reliability-aware spectral framework for federated OCT classification under client-dependent annotation noise and heterogeneous data distributions. OCT-FedSIR combines class-balanced spectral estimation, Stage-I logit adjustment, complementary spectral descriptors, selective spectral relabeling, and noise-aware federated optimization. We evaluated the framework on the Kermany, University of Illinois Chicago, and Wake Forest datasets under symmetric and structured asymmetric noise and three levels of non-IID heterogeneity. Across 117 experimental conditions, OCT-FedSIR achieved a mean accuracy of 86.73%, compared with 79.94% for RoFL and 78.75% for FedCorr. It correctly separated clients with original and corrupted annotations across all evaluated conditions, while the original FedSIR identification procedure was less robust, particularly under asymmetric noise. Spectral relabeling recovered 77.2% of corrupted annotations with 91.3% correction precision and a 3.5% false-correction rate. Retaining corrected clients outperformed spectral pruning by 9.30 percentage points on average. These findings show that annotation noise can often be identified and corrected without discarding informative client data.